How measurements work
A frequency response curve looks like a clean, objective number. In reality, every curve you see is the result of several choices — and if you compare curves that were made with different choices, you can end up thinking two headphones sound different when really the rigs sound different.
This page is a brief, non-technical tour of where a curve actually comes from, so the rest of the guide makes sense.
Measurements = headphone + α
Imagine you put a headphone on a mannequin head, the head has a microphone where an eardrum would be, and you play a test signal through the headphone. The microphone records what reaches the "eardrum". A computer turns that recording into the bumpy line you see on the graph.
So the line on screen isn't "the headphone". It's:
What this particular headphone sounds like
through this particular rig as recorded by this particular method
The curve you'll typically see in modernGraphTool is raw: it's the microphone's recording, in SPL, plotted against frequency. No correction, no compensation whatsoever. Most measurement sites — including most CrinGraph-era squig.link sites — default to this raw view. That's the data you're looking at.
Change the headphone, the curve changes. Change the rig, the curve also changes, even if the headphone is identical. That's not a flaw — it's how acoustic measurement works — but it's why you need to know a little about where a curve came from before comparing curves across sites.
Couplers and rigs
The "fake ear" — usually just the simulated ear canal and eardrum — is technically called a coupler. When you add a rubber outer ear and mount it on a stand, reviewers casually call the whole setup a rig. If it's shaped like a whole human head and torso, it's a HATS (Head and Torso Simulator). Whatever the shape, it's the thing that catches the sound coming out of the headphone and converts it into a signal the measurement software can look at.
Different labs use different couplers. You'll see names like IEC 60318-4 (often shortened to "711"), B&K 5128, GRAS KEMAR, and a few others. Each one is built differently — different ear canal shape, different eardrum simulator, different microphone — and each one hears the same headphone slightly differently. That's physics, and whatever they pick up goes straight into the graph as raw SPL.
Two sites measuring the same headphone on two different rigs will show two different raw curves, and neither of them is wrong. They're just measurements on different instruments.
So how do you read a raw curve?
Without any correction, a raw curve looks nothing like a flat line — it has the rig's own resonances baked in, plus whatever ear-canal amplification the coupler models. On a 711 rig most headphones will show a big peak around 3 kHz and a second rise up near 8 kHz, because the rig does that, not because every headphone in existence has exactly the same treble.
You make sense of that shape by overlaying a target that was designed for the same rig. A target is a reference line showing the raw shape a headphone is expected to measure on that rig in order to match some chosen definition of "good sound" — and the choice depends on which target you pick.
Some targets represent neutrality (what a neutral-sounding headphone would measure). Others represent research-derived listener preference with deliberate bass or treble shaping (Harman). Others represent a single reviewer's taste. All three are reference lines you can overlay, and all three are encoded so that matching the target means matching whatever it represents — not always "neutral". Page 04 · Why targets exist goes into the philosophies in detail.
Two practical consequences:
- Use the target that matches the rig. The operator of each measurement site picks targets meant for the rig they use. A target drawn for a 711 rig is different from a target drawn for a B&K 5128 rig. Apply the wrong one and the comparison is meaningless — you'd be reading the rig mismatch as a tuning flaw.
- The target is how you minimize the rig's influence. Without it, a raw curve is just "what the mic heard". With it, the vertical distance between the headphone and the target tells you how the headphone deviates from whatever reference you chose.
Some measurement sites (less common) pre-compensate the curve itself — subtracting the rig's house shape before drawing, so the resulting graph is a "corrected" curve that hovers near a flat line when the headphone hits whatever reference the site uses. That's a different display style, not a different measurement. modernGraphTool's default assumption, matching what most squig.link sites do, is that you're looking at raw curves and overlaying a matching target.
Sample-to-sample variation
Even if you stick to one site and one rig, measurements are still noisy. Put the same headphone on the same head twice and you'll get two almost identical curves, not identical ones. The fit, the seal against the ear, tiny re-seatings of the cable — they all nudge the result a little.
That's why some devices in the tool show a shaded band around their curve instead of a single sharp line. The band represents the spread between multiple measurements of that headphone (either multiple units, or the same unit measured multiple times). A narrow band means "we're confident this is what it does". A wide band in a particular region means "the result there depends on how it fits".
You can toggle the shading style, and you can also view each individual sample as its own curve — the per-curve controls are covered in Working with curves.
So how do I compare anything?
Three rules of thumb, in decreasing order of how safe they are:
- Comparing two curves from the same site is the most trustworthy comparison. Same rig, same operator. Differences between two curves here are very likely real differences between the headphones, and the site's target will be correct for both.
- Comparing curves from different sites that use the same rig is usually fine. Not identical — different operators, different seating — but close enough to reason about, and you can use either site's rig-matching target.
- Comparing curves from different rigs is risky. The shapes can look dramatically different even when the headphones sound similar in reality, and a target that fits one rig won't fit the other. If you must, focus on the overall tilt and on whether there's a big dip or peak in the same region, rather than on small dB differences.
modernGraphTool is designed for site operators to configure targets that match their rig, which is another reason each deployment can ship a different target list. See Why targets exist for what targets are actually doing in all of this.
Measurement science is a much bigger topic than one page can cover — real acoustics literature handles diffraction, ear canal anatomy, probe mic placement, and a dozen other things this page glossed over. The goal here is just to give you enough mental model to not be surprised when two graphs disagree.